Thursday, August 11, 2005

Redacting Social Security Numbers

The software I saw demonstrated on Monday (aiIndex) has another very useful application. It can search through scanned document images and located social security numbers. It does this by looking for number combinations and the types of forms most likely to have such numbers. Once it finds a possible social security number, it highlights it so the computer operator can decide whether to redact it (black it out) or not. A registry in another state did a test in which employees looked through a certain number of records to identify all of the instances where social security numbers appeared, then the registry sent the same stack of documents to the designers of aiIndex and let them run their software on it. Well, the computer found 30% more social security numbers than the human beings did when searching the same documents. A very nice feature of this program is that the original page that contains the social security number is saved by the computer in a safe place and the exact location of the "blacked out" part is also stored so that if the change ever has to be undone (if for example, after blacking out entire social security numbers, the law requires only the first five digits to be removed). We haven't yet asked for a cost estimate to obtain this software. I suspect it's pretty high. But if the state legislature demands that we remove social security numbers from our records (there are several bills pending that would require this) we will need a product such as this to make that requirement happen.

No comments: